Frank E Harrell Jr wrote:
>
> [Note: The original post should have been to sci.stat.consult, not sci.stat.edu]
>
> Categorizing continuous variables to avoid a linearity assumption is always a
>curious thing to do in my view. Is a piecewise flat relationship more realistic than
>a linear one?
>
> Regression splines and other flexible approaches do away with the need for the
>linearity assumption anyway.
Well, yeah, I would say so. If the relationship is U-shaped, say, then
re-coding a continuous predictor variable into 5 categories, will provide a
better fit.
I think, too, that one reason for preferring categorical predictors, at least in
my field, nutritional epidemiology, is that the shape of the relatinship may be
unknown a priori, and using categories results in a more flexible model.
-Jay
.
.
=================================================================
Instructions for joining and leaving this list, remarks about the
problem of INAPPROPRIATE MESSAGES, and archives are available at:
. http://jse.stat.ncsu.edu/ .
=================================================================
